Abstract
Wireless networks supporting artificial intelligence have gained significantattention, with Over-the-Air Federated Learning emerging as a key applicationdue to its unique transmission and distributed computing characteristics. Thispaper derives error bounds for Over-the-Air Federated Learning in a Cell-freeMIMO system and formulates an optimization problem to minimize optimality gapvia joint optimization of power control and beamforming. We introduce theMOP-LOFPC algorithm, which employs Lyapunov optimization to decouple long-termconstraints across rounds while requiring only causal channel stateinformation. Experimental results demonstrate that MOP-LOFPC achieves a betterand more flexible trade-off between the model's training loss and adherence tolong-term power constraints compared to existing baselines.